getCurrentState
Serializes the current state of the learning model for persistent storage.
This method captures the complete state of the learning engine including model weights, hyperparameters, training history, and configuration settings. The serialized state enables seamless continuation of learning across application sessions and provides the foundation for backup, recovery, and migration scenarios.
State components captured:
Model weights and biases: All trainable parameters of the neural network or ML model
Hyperparameters: Learning rates, regularization settings, architectural parameters
Training metadata: Learning progress, iteration counts, convergence metrics
Configuration settings: Model architecture definitions, feature engineering parameters
Historical statistics: Performance metrics, validation scores, adaptation indicators
Serialization requirements:
Completeness: All information needed to restore identical model behavior
Versioning: Include format version for backward compatibility and migration
Compression: Efficient encoding to minimize storage space and transfer time
Integrity: Checksums or validation data to detect corruption
Privacy: Ensure no sensitive user data is inadvertently included
Return
A KarlContainerState object containing the serialized model state, metadata, and version information required for restoration in future sessions.
See also
for state restoration process
for state format documentation
Throws
if engine is not properly initialized
if model state cannot be properly serialized